Imagine a team celebrating a quick first-reply number. Every request receives “We are looking into it” before lunch. By Friday, several customers are still waiting for the answer. The first number is real. It simply does not describe the whole service.
Useful customer support metrics show how quickly people receive a meaningful response, how long work stays open and whether the answer lasts. Pair timings with a review of actual conversations. Do not let a tidy dashboard replace reading the request.
The idea, at a glance
Read the number, then read the work.
- 01Define
State the event, clock and exclusions.
- 02Pair
Show completed work beside open work.
- 03Inspect
Read the cases behind unusual waits.
- 04Improve
Assign one change and a review date.
Define the clock before comparing the number
Write down when a clock starts, what stops it and whether it counts closed hours. First reply and resolution measure different events. Zendesk's definitions distinguish first reply, first resolution and latest resolution, as well as business-hours measures.
In your own report, state whether an automatic acknowledgement is excluded from a meaningful first response. Do not compare that local measure with a provider's figure unless the definitions match. Keep unresolved and unanswered work visible rather than leaving it out because it has no completion time.
Begin with a small weekly review
| Question | Measure to inspect | Check beside it |
|---|---|---|
| Are new requests noticed? | Time to first useful reply | Requests still unanswered |
| Does work finish? | Time to resolution under a stated rule | Age of requests still open |
| Are promises kept? | Due updates completed on time | Which overdue items need an owner |
| Do answers hold up? | Requests reopened for the same problem | Whether closure was premature |
| What repeats? | Requests grouped by reason | One process or article to improve |
These are suggested review questions, not universal targets. Use periods that match your volume. A team with a handful of requests should read the cases before drawing conclusions from a percentage.
Calculate a typical wait without hiding the long one
Compare the middle with the average
Consider five invented first-reply times: 10, 15, 20, 25 and 180 minutes. Their sum is 250 minutes. The average is 250 divided by 5, or 50 minutes. The median, the middle value after sorting, is 20 minutes.
Both figures are correct. The 180-minute request still matters, so show the longest waits and inspect the reason. Neither the average nor the median can tell you whether the actual answer helped.
Keep unfinished work in the discussion
Imagine ten requests arrive in a week and eight finish. Reporting only those eight can make resolution look quick while the oldest two continue to wait. Show the count and age of open requests beside completed work. A change in request mix can explain a change in timings without anyone becoming better or worse.
Read examples before judging people
Review a routine request, a complicated one and a reopened one together. Ask whether the owner had the authority, information and staffing needed. Avoid a league table that rewards easy work and makes colleagues reluctant to take difficult cases.
The Ombudsman's complaint principles treat learning from complaints as part of improving service. Although written for UK public bodies, that review habit is useful here without adopting a legal standard or a performance benchmark.
Turn one finding into a named change
If updates go overdue during leave, try the handover checklist. If similar questions reopen, inspect the help article. If urgent cases are buried, review the priority rules.
Write the change, owner, review date and expected sign of improvement. Compare similar periods and note holidays, staffing changes and unusual cases. A weekly review earns its time when it leads to a specific action someone can complete.
A conversation worth having
Reader comments
What has worked in your business? Share a useful example or ask a question. Comments are reviewed before publication.
Loading comments…